Add
  - CutOffKernel which uses the matrixkernel class in the P2P Operator
  - CutOffAlgorithm to explain how we can use scalFMM with only cutOff method. The Octree dos not contain any local or multipole values.

Major modification! Defined rigoreous target-source convention: p(x)=sum_y K(x,y)w(y) with x=target and y=source. Direct computation is done by calling MutualParticles(Target params,Source params) or P2P(Target, Source); Kernel matrices evaluation is done by calling evaluate(x=target,y=source) which returns k(x,y)=k(x-y).

Removed blas interface inclusion in matrix kernel header since blas not actually used here; consequently removed FUSE_BLAS in examples/DirectComputation.

Provided generic functions for the evaluation of the direct interactions (pass matrix kernel as argument).

FmaR4W8Particle does not exist anymore, changes to the only class using it : DirectComputation, done